Executive Summary
HKUST researchers' SkillCloak shows simple packing changes let malicious AI agent skills evade static scanners over 90% of the time; they also built a runtime checker.
What Happened
Scanners meant to catch malicious add-on "skills" for AI coding agents can be fooled by a few simple changes that leave the malware working, according to a new study from researchers at the Hong Kong University of Science and Technology. Their strongest trick slipped past every scanner tested more than 90% of the time, and the same team built a runtime checker that catches most of the
Vulnerability / Attack Technique
Evasion of AI skill scanners
Technical Mechanism
As reported, packed skills hide malicious content from static analysis.
AI Security Category
LLM03:2025 Supply Chain
Part of the AI Supply Chain and AI SBOM topic.
Relevant Frameworks
RealGround Analysis
Static skill scanning is not enough. RealGround analysis: combine allowlists with runtime monitoring.
Recommended Controls
- Allowlist skill publishers
- Monitor skill behaviour at runtime
Related RealGround Service
AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/07/new-skillcloak-technique-lets-malicious.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.